Upper extremity dysfunction is an important part of the disability caused by stroke, the leading cause of long term disability in adults. An evolving body of research suggests that specific rehabilitation interventions can improve functioning in an impaired arm even after sustained loss of function. We and others have shown that robot assisted rehabilitation can reduce arm impairment compared to baseline. We have now shown that robot assisted rehabilitation can reduce arm impairment in a randomized, single-blind trial compared to conventional occupational therapy. Studies of robotic training and also constraint-induced therapy suggest that specific training on functional tasks may enhance to translation of training gains into improved function. We propose to study the value of adding a functional training session (which we refer to as transition to task training or TTT) to robotic training. One published study suggests that robotic training induces changes in cortical motor maps. We will conduct studies of cortical motor excitability in our patients before and after training to investigate this issue. Criteria for determining the maximum benefit of robotic training have not yet been developed. One approach would be to conduct conventional upper extremity evaluations during the course of therapy but that would significantly increase therapist time. The robots record the assistance provided to patients during each training session and we will examine whether that data might be used on an ongoing basis to evaluate patient benefit. Primary hypothesis: TTT immediately following each robotic training session leads to greater functional improvement than robot- assisted training alone for patients with chronic upper extremity dysfunction due to hemi-paretic stroke. Secondary mechanistic hypothesis: Robotic training results in alterations in cortical motor excitability as detected by transcranial magnetic stimulation. Secondary clinical hypothesis: Robot kinematic data will reflect the functional changes seen in the training and that this information can be used clinically to determine when a patient has reached maximum benefit from robotic training. In order to test these hypotheses we propose a randomized, single-blind trial comparing 36-60 minute robotic training sessions with 36 training session consisting of 45 minutes of robotic training and 15 minutes of TTT training. Outcomes will include upper arm functional assessments, the results of transcranial magnetic stimulation and robot generated kinematic data. Page 1